<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=ISO-8859-7"
http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Dmitry,<br>
<br>
thanks for the interesting results.<br>
<br>
Your results are not as promising as mine.<br>
I guess, referring to the HotSpot's dissassembler dump would give more
hints.<br>
Additionally different string lengths, repeat factors, -Xbatch, and GC
settings would help.<br>
<br>
Good trick to use substring(1) for cheap instatiation.<br>
<br>
To estimate the instantiation cost you could add:<br>
public int timeInitLong(int reps) {<br>
int dummy = 0;<br>
for (int i = 0; i < reps; i++) {<br>
dummy += LONG.substring(1).offset - 1;<br>
}<br>
return dummy;<br>
}<br>
<br>
Do you have any idea, why first hashCode1() warmup is 10 times slower
than hashCode() warmup in my case?<br>
<br>
-Ulf<br>
<br>
<br>
Am 27.02.2010 15:00, schrieb Dmitry Nadezhin:
<blockquote
cite="mid:5a8e758b1002270600j71a9dff8t84bc4cf0262f5032@mail.gmail.com"
type="cite">Ulf,<br>
<br>
I ran Caliper benchmarks on different variants of hashCode().<br>
<a moz-do-not-send="true" href="http://code.google.com/p/caliper/">http://code.google.com/p/caliper/</a><br>
The results and the Caliper code are below:<br>
<br>
-Dima<br>
<br>
<br>
<div class="gmail_quote">On Sat, Feb 27, 2010 at 2:18 AM, Ulf Zibis <span
dir="ltr"><<a moz-do-not-send="true" href="mailto:Ulf.Zibis@gmx.de">Ulf.Zibis@gmx.de</a>></span>
wrote:<br>
<blockquote class="gmail_quote"
style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="im">Am 26.02.2010 20:32, schrieb Alan Bateman:<br>
</div>
<div class="im">
<blockquote class="gmail_quote"
style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">Ulf
Zibis wrote:<br>
For these other suggestions it would be great to create<br>
micro-benchmarks and try our your changes.<br>
<br>
For now though, as I said, I'm just fixing the method to avoid the<br>
store for the empty string case.<br>
</blockquote>
<br>
</div>
Here the benchmark for my different approaches on non-empty strings:<br>
<br>
hashCode() warmup = 38,227 ms<br>
hashCode1() warmup = 323,462 ms<br>
hashCode() time = 12,066 ms<br>
hashCode1() time = 11,620 ms<br>
hashCodes sum = 0<br>
hashCode() warmup = 34,928 ms<br>
hashCode1() warmup = 19,010 ms<br>
hashCode2() warmup = 25,938 ms<br>
hashCode3() warmup = 22,428 ms<br>
hashCode4() warmup = 22,966 ms<br>
hashCode5() warmup = 21,372 ms<br>
hashCode() time = 418,962 ms<br>
hashCode1() time = 306,977 ms<br>
hashCode2() time = 296,899 ms<br>
hashCode3() time = 363,364 ms<br>
hashCode4() time = 286,098 ms<br>
hashCode5() time = 269,519 ms<br>
hashCodes sum = -1427881024<br>
<br>
See new revision:<br>
<a moz-do-not-send="true"
href="https://java-nio-charset-enhanced.dev.java.net/source/browse/java-nio-charset-enhanced/branches/j7_EUC_TW/src/java/lang/String.java?rev=955&view=markup"
target="_blank">https://java-nio-charset-enhanced.dev.java.net/source/browse/java-nio-charset-enhanced/branches/j7_EUC_TW/src/java/lang/String.java?rev=955&view=markup</a><br>
<font color="#888888">
<br>
-Ulf<br>
<br>
<br>
<br>
<br>
</font></blockquote>
</div>
<br>
</blockquote>
</body>
</html>